The role of Artificial Intelligence in healthcare: potential use in national clinical audit

Published: 21 Nov 2023

Dr Claudia Snudden, Clinical Fellow and Drew Smith, Associate Director at HQIP

In the ever-evolving realm of technological advancements, Artificial Intelligence (AI) stands out as a beacon of transformative change. From smart home systems to predictive algorithms in finance, AI is reaching into more and more corners of our modern existence. The face of this revolution is generative AI, that is: learning models that can generate text, images, and other content based on the data they were trained on. Perhaps the most well known generative AI tool is OpenAI’s ChatGPT, but others such as Bing and Bard work on the same principle. This rapid assimilation of data and generation of information is not just transforming industries on a macro scale; it’s redefining tasks as granular as administrative chores and as expansive as biomedical research.

For those of us at The Healthcare Quality Improvement Partnership (HQIP) and our audit providers, the rise of AI offers both challenges and opportunities. We stand at the cusp of a new era where we must consider the nuances of integrating this technology into our daily workflows. How can we harness it for efficiency, while also navigating its complexities to ensure we remain compliant, especially concerning data protection? In this exploration, we’ll delve into two ways that generative AI can add value to national clinical audits: data analysis and content generation.

Your virtual data analyst

The processing and interpretation of data can often be a bottleneck for many organisations, especially in the realm of healthcare where accuracy is paramount. Generative AI can alleviate some of these challenges. For programmers and analysts, AI can act like an on-demand mentor, providing insights on programming languages like Python, Java, or R, offering solutions to coding conundrums, and even guiding on the nitty-gritty of debugging. It can also advise on data cleaning techniques, suggest appropriate statistical tests, and recommend variables or relationships worthy of investigation.

By feeding or uploading aggregate non-personalised data to the model, you can receive not just analyses but also guidance on presentation. Should you opt for a bar chart or a scatter plot? Which patterns should be highlighted for your target audience? Even more excitingly, as the functionality of AI tools grows these visualisations can be crafted in real-time, serving as a dynamic aid in data interpretation and communication.

Content creation and synthesis

In an age of information overload, the ability to effectively communicate findings is as critical as the discoveries themselves. Beyond numbers and codes, generative AI boasts impressive capabilities in the domain of text. The myriad of possibilities here are genuinely extensive: from distilling the core findings of a lengthy research paper into a concise summary, to translating content across languages, or to converting raw bullet points into coherent prose.

And it’s not just about simplification; it’s about tailoring. An audit report, for instance, could be transformed into an accessible blog post or an informative pamphlet, depending on the target audience. Stuck with where to start when writing a report? Upload your outline to a generative AI tool and let it kick off the creative process for you.

Walking with caution: the limitations

As with any technological advancement, generative AI isn’t without its limitations. A primary concern resides in the realm of data protection and privacy. It’s vital to approach online tools as if you’re releasing information into the public domain. Personalised or even pseudonymised details should never be entered into such a tool. Similarly, one should refrain from uploading unpublished content such as national clinical audit report drafts.

Moreover, whilst the reservoirs of knowledge that AI tools are built on are vast they are not infinite, nor are they always up-to-date. Tools may sometimes stumble with deeply technical jargon or misinterpret the context behind a query.

With all this in mind our most important caveat is that artificial intelligence is… well… artificial. As exciting, insightful, and revolutionary as AI can be, it doesn’t (yet) replace the 2 million years of human evolution that’s gone into us. Generative AI tools are known to ‘hallucinate’, that is: to create incorrect information and present it as fact. Even, in some cases, fabricating its own citations. AI hallucinations can be difficult to spot and arise from a model being trained on a limited, out-of-date, or low-quality data set. Or they can occur because the tool doesn’t possess the contextual understanding of the prompt.

It follows therefore that as users the onus is on us to cross-check citations provided by generative AI, and sense-check the information it produces. As Gartner points out: “Generative AI creates artifacts that can be inaccurate or biased, making human validation essential and potentially limiting the time it saves workers.”

Looking forward

The incorporation of generative AI within both the infrastructure at HQIP and our audit providers promises a slew of benefits. From data analysts to communication specialists, the potential for AI models to elevate our workflows, bolster efficiency, and sharpen our outputs is undeniable. As we look ahead, there’s also the exhilarating prospect of using tools to develop bespoke AI analytical models tailored for audit data extraction and analysis across clinical audit themes. This endeavour, though filled with promise, will require concerted effort, continual learning, and strategic investments.

Current models, whilst impressive, largely result in users being able to perform existing tasks more quickly. But in the next few years we anticipate that continued growth in investment and computational power will see a blossoming of AI capabilities and an increasing democratisation in the use of tools.

The advent of generative AI offers a glimpse into a future brimming with possibilities. As we embark on this journey, let’s tread with curiosity, caution, and the unwavering aim to better our practices for the greater good. The future of HQIP, in many ways, is intertwined with the possibilities this technology brings, and it’s a future we eagerly anticipate.